Defects in semiconductor manufacturing processes result in lost production and fabrication downtime. The cause of these defective products is hard to determine as today's manufacturing processes involve not only a single manufacturing operation but multiple instances of the same operation performed by different machines and a multitude of other processes and machines. Identifying the cause of the error is in itself time-consuming and the large amounts of data to examine makes quick and accurate analysis difficult.
During the semiconductor manufacturing process the unprocessed wafer proceeds through several distinct manufacturing processes. Measurements are taken during this process and analyzed. If the analysis shows that a problem is occurring the process is stopped and the condition resolved before production is allowed to resume. These processes can become very complicated and data analysis may not be able to be completed and studied until well after the time that the particular wafer that had been processed has been completely processed by the faulty process.
Further compounding the lag time between the measurement and the analysis and then the subsequent reaction is that for a single process, a number of machines may perform that same exact process. The number of measurements taken during this manufacturing process expands linearly and data analysis is no longer cumbersome, it is impossible. Anomalies in the processes are not identified quickly enough and entire lots of end-product may be defective as a result.
The problem is that there is no easy way to combine the data and reduce the data processing time so meaningful data analysis can take place and reaction to current conditions can take place quicker. What is needed is a way to aggregate processes and resulting data for the purposes of analysis.